Author Affiliations
Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence, College of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, Chinashow less
Fig. 1. Framework of stereo image quality assessment algorithm based on fused image and convolutional neural network
Fig. 2. Left and right views and fused image of stereo image. (a) Left view; (b) right view; (c) fused image
Fig. 3. Results with normalized luminance coefficient. (a) Fused image with normalized luminance coefficient;(b) probability density distribution
Fig. 4. Architecture of convolutional neural network
Fig. 5. Predicted scatter plots. (a) LIVE 3D Phase I database; (b) LIVE 3D Phase II database
Method | RSROCC | RPLCC |
---|
LP | 0.9462 | 0.9687 | NSCT | 0.9455 | 0.9688 | GFF | 0.9483 | 0.9642 | DWT | 0.9429 | 0.9675 |
|
Table 1. RSROCC and RPLCC of different fused methods on LIVE 3D Phase I database
Method | RSROCC | RPLCC |
---|
LP | 0.9224 | 0.9237 | NSCT | 0.9218 | 0.9243 | GFF | 0.8955 | 0.9072 | DWT | 0.9342 | 0.9428 |
|
Table 2. RSROCC and RPLCC of different fused methods on LIVE 3D Phase II database
Method | RSROCC | RPLCC |
---|
LP | 0.7940 | 0.7763 | NSCT | 0.7828 | 0.7765 | GFF | 0.8012 | 0.7921 | DWT | 0.8033 | 0.7822 |
|
Table 3. RSROCC and RPLCC of different fused methods on MCL-3D database
Method | JP2K | JPEG | WN | BLUR | FF | All |
---|
PSNR | 0.799 | 0.121 | 0.932 | 0.902 | 0.587 | 0.834 | SSIM | 0.858 | 0.436 | 0.938 | 0.879 | 0.586 | 0.876 | Method in Ref. [1] | 0.910 | 0.603 | 0.929 | 0.931 | 0.699 | 0.899 | Method in Ref. [10] | 0.875 | 0.614 | 0.943 | 0.937 | 0.781 | 0.921 | Method in Ref. [9] | 0.866 | 0.675 | 0.914 | 0.555 | 0.640 | 0.383 | Method in Ref. [8] | 0.863 | 0.617 | 0.919 | 0.877 | 0.652 | 0.891 | Method in Ref. [15] | 0.889 | 0.613 | 0.909 | 0.877 | 0.758 | 0.925 | Method in Ref. [14] | 0.837 | 0.638 | 0.931 | 0.833 | 0.649 | 0.892 | Method in Ref. [16] | 0.917 | 0.782 | 0.910 | 0.865 | 0.666 | 0.911 | Proposed | 0.786 | 0.929 | 0.952 | 0.964 | 1.000 | 0.943 |
|
Table 4. RSROCC of different methods on LIVE 3D Phase I database
Method | JP2K | JPEG | WN | BLUR | FF | All |
---|
PSNR | 0.785 | 0.219 | 0.935 | 0.916 | 0.703 | 0.834 | SSIM | 0.865 | 0.487 | 0.939 | 0.919 | 0.721 | 0.872 | Method in Ref. [1] | 0.940 | 0.641 | 0.925 | 0.949 | 0.747 | 0.902 | Method in Ref. [10] | 0.923 | 0.656 | 0.941 | 0.951 | 0.840 | 0.924 | Method in Ref. [9] | 0.905 | 0.729 | 0.904 | 0.617 | 0.503 | 0.626 | Method in Ref. [8] | 0.907 | 0.695 | 0.917 | 0.968 | 0.735 | 0.895 | Method in Ref. [15] | 0.898 | 0.632 | 0.923 | 0.928 | 0.845 | 0.926 | Method in Ref. [14] | 0.848 | 0.626 | 0.925 | 0.899 | 0.707 | 0.887 | Method in Ref. [16] | 0.938 | 0.806 | 0.919 | 0.881 | 0.758 | 0.917 | Proposed | 0.984 | 0.870 | 0.968 | 0.985 | 0.991 | 0.968 |
|
Table 5. RPLCC of different methods on LIVE 3D Phase I database
Method | JP2K | JPEG | WN | BLUR | FF | All |
---|
PSNR | 0.597 | 0.491 | 0.919 | 0.690 | 0.730 | 0.665 | SSIM | 0.704 | 0.678 | 0.922 | 0.828 | 0.834 | 0.792 | Method in Ref. [1] | 0.751 | 0.867 | 0.923 | 0.455 | 0.773 | 0.728 | Method in Ref. [10] | 0.847 | 0.719 | 0.845 | 0.800 | 0.850 | 0.745 | Method in Ref. [9] | 0.724 | 0.649 | 0.714 | 0.682 | 0.559 | 0.543 | Method in Ref. [8] | 0.867 | 0.867 | 0.950 | 0.900 | 0.933 | 0.880 | Method in Ref. [14] | 0.553 | 0.593 | 0.593 | 0.869 | 0.828 | 0.825 | Method in Ref. [16] | 0.864 | 0.839 | 0.932 | 0.846 | 0.860 | 0.888 | Proposed | 0.929 | 1.000 | 0.964 | 0.964 | 0.929 | 0.934 |
|
Table 6. RSROCC of different methods on LIVE 3D Phase II database
Method | JP2K | JPEG | WN | BLUR | FF | All |
---|
PSNR | 0.597 | 0.491 | 0.919 | 0.690 | 0.730 | 0.665 | SSIM | 0.704 | 0.678 | 0.922 | 0.838 | 0.834 | 0.792 | Method in Ref. [1] | 0.784 | 0.853 | 0.926 | 0.535 | 0.807 | 0.748 | Method in Ref. [10] | 0.837 | 0.750 | 0.849 | 0.827 | 0.880 | 0.758 | Method in Ref. [9] | 0.766 | 0.786 | 0.722 | 0.795 | 0.674 | 0.568 | Method in Ref. [8] | 0.899 | 0.901 | 0.947 | 0.940 | 0.932 | 0.880 | Method in Ref. [14] | 0.634 | 0.647 | 0.904 | 0.967 | 0.851 | 0.818 | Method in Ref. [16] | 0.867 | 0.829 | 0.920 | 0.878 | 0.836 | 0.845 | Proposed | 0.939 | 0.932 | 0.971 | 0.987 | 0.960 | 0.943 |
|
Table 7. RPLCC of different methods on LIVE 3D Phase II database
Metric | 3DSwIM | You(l) | You(g) | StSd | Benoit | Gorley | ST-SIAQ | NIQSV | Proposed |
---|
RPLCC | 0.6497 | 0.7504 | 0.3650 | 0.6995 | 0.7425 | 0.7099 | 0.7133 | 0.6783 | 0.7822 | RSROCC | 0.5683 | 0.7567 | 0.6609 | 0.7008 | 0.7518 | 0.7196 | 0.7034 | 0.6208 | 0.8033 |
|
Table 8. RPLCC and RSROCC of different methods on MCL-3D database